For a system which is required to be highly reliable and available and is not allowed to cause communication timeout while constantly communicating with clients, a dual computer as shown in Patent Document 1 is used. For example, a dual system is applied to an application server, a virtual PC (Personal Computer) server of a thin client system, a production management server used in a manufacturing factory, an online payment system, and so on.
A computer that configures such a dual system has major hardware components in twos within the system, such as a CPU (Central Processing Unit), a memory and a PCI (Peripheral Component Interconnect) adapter, and the components are caused to synchronize by a dedicated LSI (Large Scale Integration) and a driver. Thus, an OS (Operating System) of the dual system operates regardless of a dual state of the hardware.
When one of the hardware components causes an error, the dual system isolates the hardware component from the system and temporarily continues to operate with only the other hardware component (this state is referred to as a fallback state). After that, the isolated hardware component is initialized and installed into the system again. If the installation is completed without any problem, the system returns to the dual state.
[Patent Document 1] Japanese Unexamined Patent Application Publication No. JP-A 2009-245015
Although the application and the OS operate without any specific problem when the system is in the dual state as described above, the system may be affected when the hardware component is installed into the system. To be specific, when the CPU is installed, memory copy from the main system to the sub system is performed so that the memories of both the main system and the sub system are caused to synchronize. A memory region that is constantly used by the system and always changed may be copied while the operation of the system is temporarily stopped. As a result, there is a case where communication with other computers is interrupted during the copy process and communication timeout occurs on a connected client computer that the system provides with an information processing service, and there is a possibility that execution of the application program is in trouble. A general system can avoid timeout by regulating TCP timeout values on both the server computer and the client computer, but the dual system is applied to a system required to be reliable and available because of the characteristics of the dual system. Therefore, it is desirable that only the server deals with the trouble.